You make most of a good review before the meeting starts. The prerequisite most teams skip is a career framework: a written, shared bar that says, in concrete behaviours, what "good" looks like at each role and level. With that in place, preparing and running a review is a repeatable flow: have the person write a structured self-assessment, gather targeted feedback from the juniors, peers and seniors they worked with, synthesise it against the framework, calibrate with the other leads, then deliver a forward-looking message that ends in a real development and staffing plan, with no surprises.
The meeting itself creates little of a review's quality. After 100 years of research, there is little evidence that appraisal on its own improves performance (DeNisi & Murphy, Journal of Applied Psychology, 2017). Feedback can also cut the wrong way: across 607 studies it raised performance on average, but over a third of feedback interventions made performance worse (Kluger & DeNisi, Psychological Bulletin, 1996). The manager decides which way it goes. Gallup attributes at least 70% of the variance in team engagement to the manager (Gallup, 2015).
Why do most team reviews go wrong before the meeting?
They rest on one manager's memory, assembled the night before. When the inputs are thin, the meeting can't save it. A poorly delivered review can set the person back, which is what Kluger and DeNisi found in more than a third of cases (1996). The fix is better preparation feeding the room.
That preparation does two jobs. It replaces "what I remember" with evidence from the people who saw the work, and it gives the conversation a forward direction so something changes afterwards. That is the only condition under which reviews reliably move performance (Smither, London & Reilly, Personnel Psychology, 2005). Get the prep right and the meeting becomes short and calm. Skip it and no amount of warmth in the room will make the verdict trusted.
What has to exist before you run a review? A career framework
Before anyone writes a word, the team needs a career framework: each role, at each level, described in about four or five concrete, behaviour-anchored bullets. Write something a junior could read and recognise, instead of "demonstrates strategic impact". A junior in client meetings, for example, listens attentively, takes notes, contributes when asked, and asks questions when something is unclear. The framework covers both the craft of each role (design, engineering, analysis) and the shared dimensions every level needs (leadership, communication, problem-solving), and the exact bar can differ from one firm to the next. This is the practitioner standard: a framework is a map of the behaviours valued and recognised at each level (CIPD, Competence and competency frameworks).
Two rules make it work. Agree it with the leads first: every manager who will rate against it has to accept the same bar, or your "Grade 3" and mine mean different things. Then share it openly with the whole team, so expectations are public instead of a secret unveiled at review time. A person can only self-assess honestly against a bar they can see.
How do you prepare and run the review, step by step?
Eight steps take you from the framework to a delivered, forward-looking review.
- Start from the shared bar. Pull the framework's role-and-level definition for this person. You measure everything that follows against it, and leave your mood and the last engagement out of it.
- Have the person write a four-part self-assessment. (1) Their biggest wins for the period: craft, organisational, personal. (2) Which of their skills drove those wins. (3) A self-rating on each dimension on a clear four- or five-point scale, with a short reason for each rating. (4) What support they need from the firm to grow. A reason per rating turns a number into a conversation.
- Ask them to name 3–5 close colleagues across levels. People they worked with, including at least one junior, one peer and one senior, so the picture comes from more than a single vantage point. On a staffed team, those are rarely the same people you'd pick.
- Collect the feedback: gather first, judge second. Ask each named colleague for specifics tied to the framework's behaviours instead of a global score. Resist forming the verdict while you collect. You are assembling evidence at this stage, so park any hunch.
- Synthesise against the framework. Merge the self-assessment, the colleague feedback and your own observation into one read per dimension. More independent sources cancel noise that a single rater can't. The gain is modest, though, which is why the next steps matter more than the rating itself (Smither, London & Reilly, 2005).
- Calibrate with the other leads. Sit with the managers who staffed this person on other engagements and reconcile your reads against the shared bar. That step turns one lead's opinion into a rating that is fair and defensible on the partner track.
- Deliver it as a forward-looking message, with no surprises. Lead with the single thing that matters most. Tie every point to a concrete behaviour from the framework and a real example. Skip the feedback "sandwich". People see through it and lose the real message.
- Close with a plan and a date. End with a concrete development plan and staffing intentions (what to work on, on which engagements, with what support) and a date to revisit. A review that doesn't change what someone does next has failed at its main job (DeNisi & Murphy, 2017).
How much does the feedback round move performance?
Less than people hope, and that is the whole argument for steps 7 and 8. Across 24 longitudinal studies, improvement after multisource feedback was generally small, and it varied by who gave it: direct reports moved behaviour most, self-ratings not at all (Smither, London & Reilly, 2005). Keep the feedback round, and don't expect the form to work magic. Follow-up (coaching, a plan, a revisit) turns feedback into change, and collecting more ratings adds little.
What if the self-assessment and your view don't match?
Treat the gap as information to work through together. Because you gathered evidence before judging (step 4), you are comparing two reads of the same behaviours, with no need to pull rank. Put the framework on the table and walk through specific examples: "here's the bar for this level, here's what I and your colleagues saw." If their reasons surface something you missed, update your view. That's why the self-assessment carries a reason per rating. If the gap holds, name it against the evidence and move to what would close it. Authority ("because I'm the manager") ends the conversation, while evidence keeps it going.
Why does this matter for agencies and consulting boutiques?
The generic "manager fills in the form" guide assumes something you don't have: one manager who watched the person all year. In an agency or a boutique, people are staffed across several engagements under different leads, and billable pressure eats the prep time at the moment reviews come due. The manager running the review is often not the person who saw most of the work.
The prep flow exists for this reason. A shared framework lets different leads mean the same thing by a grade. A structured self-assessment surfaces wins you weren't staffed to see. A targeted junior/peer/senior round reconstructs the year from the people who were there. And calibration across leads makes the rating hold up when it feeds staffing, raises and the partner track. Get this right and the review is fair and defensible. Get it wrong and the cost lands on the senior talent a boutique can least afford to lose, because an opaque verdict reads as an unfair one. The manager carries an outsized share of that experience (Gallup, 2015).
A manager's pre-review checklist
Score your preparation: one point per "yes". Six or more means your review is ready. Four or fewer means you're improvising.
- You have a written career framework and have pulled this person's role-and-level bar.
- You agreed the framework with the other leads and shared it openly with the team.
- The person wrote a four-part self-assessment, with a reason behind each self-rating.
- They named 3–5 close colleagues spanning at least one junior, one peer and one senior.
- You collected specific, behaviour-tied feedback before forming your verdict.
- You calibrated your read with the other leads who staffed the person.
- Your delivery leads with the one thing that matters and ties every point to evidence.
- You close with a concrete development + staffing plan and a date to revisit.
FAQ
What's the single most important thing a manager can do to run a good review?
Prepare before the room. Build the verdict from a shared framework, a structured self-assessment and targeted colleague feedback. A rating delivered without that groundwork barely moves performance (DeNisi & Murphy, 2017), and a badly delivered one can make things worse in over a third of cases (Kluger & DeNisi, 1996).
Who should give feedback on someone staffed across several engagements?
The 3–5 colleagues who worked with them this period, spanning levels (at least one junior, one peer and one senior), plus the leads who staffed them on each engagement. The goal is to reconstruct the whole year, well beyond the last project.
How do I handle it when someone rates themselves much higher than I would?
Compare both reads against the framework using specific examples, and leave authority out of it. Ask for the reasons behind their self-rating. Sometimes they reveal work you missed, sometimes they reveal a real gap. Either way, keep the conversation on behaviours and evidence rather than the number.
Does collecting 360-degree feedback automatically improve performance?
No. Improvement after multisource feedback is generally small and depends on what follows it: coaching, a plan, a revisit (Smither, London & Reilly, 2005). The feedback round is the input. The development plan and follow-up create the change.
How much time should preparation take when everyone is billable?
Less than you fear, if the framework already exists and the self-assessment and colleague inputs are templated. The framework is the one-time investment. After that, each review is mostly assembling inputs you can collect asynchronously around client work.